Sample Repeatability Study
Geochemistry has always faced questions about repeatability and precision. I
undertook this project to determine the variability over time of sampling a fixed
point. The project consisted of collecting 64 samples on 4 foot centers from a
784 sq. ft. area. This tiny area ended up displaying a surprising amount of
variability. The transition from background to anomaly in this area occurs at 2.0
ppm I2, indicating this location is anomalous. The range of values is larger than
anticipated although the distribution is consistent with a single anomalous
population (see the probability plot at the bottom). One of the problems with the
project was that the collection points became small holes over the course of
collection. (Macro Heisenberg?)
undertook this project to determine the variability over time of sampling a fixed
point. The project consisted of collecting 64 samples on 4 foot centers from a
784 sq. ft. area. This tiny area ended up displaying a surprising amount of
variability. The transition from background to anomaly in this area occurs at 2.0
ppm I2, indicating this location is anomalous. The range of values is larger than
anticipated although the distribution is consistent with a single anomalous
population (see the probability plot at the bottom). One of the problems with the
project was that the collection points became small holes over the course of
collection. (Macro Heisenberg?)
The Project
This first survey demonstrated immediately significant variation over short distances in areas of seepage.
The background area, shown at the bottom of the page, has very little variation. This test proved that
integrating multiple samples from the sample area, which I have recommended since 1982, is an
important step.
The background area, shown at the bottom of the page, has very little variation. This test proved that
integrating multiple samples from the sample area, which I have recommended since 1982, is an
important step.
2/24/95
This first set, above,was taken in late February on a sunny "shirt sleeve" Colorado winter day. The
preceding months had been warmer and dryer than normal. The range is from 2.1 to 3.5 with a
mean of 2.73 and a SYD DEV of 0.29.
preceding months had been warmer and dryer than normal. The range is from 2.1 to 3.5 with a
mean of 2.73 and a SYD DEV of 0.29.
4/14/95
This second set was taken in mid April two days after a heavy 10" spring snow. The range was
from 2.4 to 3.8 with a mean of 2.92 and a STD DEV of 0.30.
from 2.4 to 3.8 with a mean of 2.92 and a STD DEV of 0.30.
8/10/95
This set was collected in the heat of the summer, near the end of the hottest and driest part of the
season. Preceding this survey, and for the first time in decades, horses grazed on this pasture. The
area of sample collection had received a liberal dose of horse waste. Some sample points were
collected after this material was removed from the collection point. The range is from 2.2 to 3.2
with a mean of 2.65 and a STD DEV of 0.28.
season. Preceding this survey, and for the first time in decades, horses grazed on this pasture. The
area of sample collection had received a liberal dose of horse waste. Some sample points were
collected after this material was removed from the collection point. The range is from 2.2 to 3.2
with a mean of 2.65 and a STD DEV of 0.28.
12/11/95
This set was collected early in the winter, not long after it had turned consistently cold. The range is
from 2.4 to 3.6 with a mean of 2.90 and a STD DEV of 0.24.
from 2.4 to 3.6 with a mean of 2.90 and a STD DEV of 0.24.
8/18/96
This last set was collected 18 months after the start of this project. The pasture was not used for
grazing this summer. The range was from 2.3 to 3.4 with a mean of 2.88 and a STD DEV of
0.21.
grazing this summer. The range was from 2.3 to 3.4 with a mean of 2.88 and a STD DEV of
0.21.
Although the range of sample values for these five sets was larger than I expected all of the
samples plot as a single population. This indicates that all 320 samples correctly identified this
area as anomalous. Single population plots are very rare. With more than a quarter of a million
samples covering hundreds of surveys I have only seen two single population plots. One in
northeastern Colorado and one in North Dakota and both were background distributions.
samples plot as a single population. This indicates that all 320 samples correctly identified this
area as anomalous. Single population plots are very rare. With more than a quarter of a million
samples covering hundreds of surveys I have only seen two single population plots. One in
northeastern Colorado and one in North Dakota and both were background distributions.
2/24/95
This background survey was done at the same time as the first survey above. The range is from
1.5 to 2.1 with none of these sample values overlapping any of the values from the surveys from
the anomalous area
1.5 to 2.1 with none of these sample values overlapping any of the values from the surveys from
the anomalous area
Conclusions
1. The most important lesson is that a method that depends on a single point collection will
produce a highly variable results from areas of seepage. Integrated samples produced by blending
soil from a number of points can eliminate some of the problems of area variability.
2. The magnitude of a sample anomaly is variable and is not as important as the identification of
the soil as anomalous.
produce a highly variable results from areas of seepage. Integrated samples produced by blending
soil from a number of points can eliminate some of the problems of area variability.
2. The magnitude of a sample anomaly is variable and is not as important as the identification of
the soil as anomalous.
GrayStone Exploration Labs, Inc.
| Research |